Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

This version of DetectorPack is compatible with Axxon One 2.0 Beta and later versions. 

What's new:

  1. Added new neural network models that simultaneously detect two classesHuman and Vehicle. The models are divided into three types: Nano, Medium and Large. Resource consumption increases as recognition accuracy increases Setting up Scene Analytics detection tools based on Neurotracker
  2. Added the definition of a "non-living face" for Face detection.
  3. Barcode detection tool.
  4. Added Face recognition VA.
  5. Added Audio analytics IVAdded the possibility of specifying the number of detection tool objects for the standard sub-detection tools: Motion in area, Appearance in area, Disappearance in area. For example, the detection tool will trigger only if two objects are moving in the frame.
  6. Added new countries for the the License plate recognition (IV) License plate recognition (IV) features and specifications
  7. Added new countries for the License plate recognition (VT) License plate recognition (VT) features and specifications
  8. Added new countries for the License plate recognition (RR) License plate recognition (RR) features and specifications

Optimization

  1. Added the possibility of quantizing all embedded models for Neurotracker Setting up Scene Analytics detection tools based on Neurotracker

  2. Unified class numbers for embedded models Setting up Scene Analytics detection tools based on Neurotracker
  3. Added the possibility of using object classes in classical sub-tools (child detection tools)Motion in area, Appearance in area, Disappearance in area.

  4. Updated the integration of the Abandoned object VIWindows version is now available.

  5. Added a new type of vehicle the cache creation utility for the —"motorcycle" for the RR vehicles recognition and License plate recognition (RR)and RR vehicles recognition Optimizing the operation .
  6. Improved the quality of speed detection for

    of

    the License plate recognition (RR)

    and RR vehicles recognition on GPU

Optimization

  1. License plate recognition (RR)

  2. Support for NVIDIA JetPack 5.1.2 (Nvidia Jetson Orin devices) in the Neuro Pack is available.
  3. Added general models for Atlas300i NPU.
  4. Normalized load by cores on Xeon processors.
  5. Improved the speed of the Image noise detection
  6. Updated the neural network algorithm OpenVINO that improved the performance up to 1.5 times when using CPU.
  7. Normalized the loading by cores on the 12th and 13th generation Intel processors and on Servers with two processors.

Update recommendations

  1. For Linux OS: if a detector pack 3.7.4 version and earlier that includes all analytics tools is installed (not divided into plugins), then before updating, it is necessary to remove the installed package without using purge (complete package deletion) by running the following command:

    Code Block
    sudo dpkg --remove --force-depends axxon-detector-pack

    and then install the necessary plugins.

  2. For Face masking to work during export, it is necessary to have the Neuro Pack installed.If you have the Move from area to area sub-detection tool created, you need to contact technical support and get a manual for updating.

Update limitations

  1. When you update the Face Recognition Pack, the results of the previously recognized persons will become unavailable. If you want to keep the previously recognized persons, do not update the analytics package.
  2. To start and correctly run the basic Face detection on GPU, you must first create a cache Optimizing the operation of Face detection on GPU
  3. Discontinued the support for the Intel HDDL devices.Removed the Move from area to area sub-detection tool from the current release to provide the support with Axxon One 1.0 software version. At the same time, the full compatibility is not guaranteed, there is no support for the declared functionality and the sub-detection tool optimization(Mustang-v-100) devices.
  4. The mechanisms for launching neural network models have changed, that is why some of the custom models (trained specifically for the object) may not work correctly. If you encounter any difficulties, please contact  technical support technical support.
  5. In this version of DetectorPack, you cannot filter objects by color when using sub-tools (child detection tools).
  6. Replaced Face recognition with Face recognition VA. Discontinued the support for Face recognition. Face recognition and Face recognition VA are incompatible.